Send
Print
Under the agreement, Anthropic will gain access to up to one million tensor processing units, also known as AI accelerators, developed by Google Cloud.
With this new agreement, Anthropic plans to deploy computing power that will surpass 1 gigawatt next year. The new chips will be used to train and support the Claude AI models, which are the foundation of the Claude AI assistant, Anthropic's main product.
Anthropic will also have access to additional Google Cloud services, which will offer advanced AI infrastructure for many years, according to the division. In addition to Google's TPUs, the startup currently uses Amazon's Trainium chips and Nvidia GPUs.
Anthropic has stated that it remains dedicated to its partnership with Amazon, its key partner in AI model training and cloud services, and will continue to collaborate with the company on Project Rainier. This is a large computing cluster that includes hundreds of thousands of AI chips spread across multiple data centers in the United States.
The startup has also mentioned its intention to continue investing in more computing resources.
source: reuters.com
With this new agreement, Anthropic plans to deploy computing power that will surpass 1 gigawatt next year. The new chips will be used to train and support the Claude AI models, which are the foundation of the Claude AI assistant, Anthropic's main product.
Anthropic will also have access to additional Google Cloud services, which will offer advanced AI infrastructure for many years, according to the division. In addition to Google's TPUs, the startup currently uses Amazon's Trainium chips and Nvidia GPUs.
Anthropic has stated that it remains dedicated to its partnership with Amazon, its key partner in AI model training and cloud services, and will continue to collaborate with the company on Project Rainier. This is a large computing cluster that includes hundreds of thousands of AI chips spread across multiple data centers in the United States.
The startup has also mentioned its intention to continue investing in more computing resources.
source: reuters.com
